Model Classification Report: precision recall f1-score support 0 0.92 1.00 0.96 33 1 1.00 0.90 0.95 31 accuracy 0.95 64 macro avg 0.96 0.95 0.95 64 weighted avg 0.96 0.95 0.95 64 This is your model when training your dataset. Mine is not like that or is it because I got the wrong data. I follow your step 1 instructions when getting data
There was similar issues with me. All i did is train it on more voice data and add more background audios with variations. And it improved with each trial.
I have been in a hurry to try this so I had launched training.py but it says missing module PCM but I have installed it using pip install pcm but again this message is coming.Any advice
@@SaloSojaEdwin Yeah in the starting I was getting many false positives too... In the part 2 I will tell how to avoid false positives almost completely! Will be uploading tomorrow or day after tomorrow...
Just record wake word with multiple words. For example, record 100 audio with first wake word and then 100 audio with the second wake word. I don't know how well this will work as I have not tried this myself.
Yeah definitely, it is a python script you can run the same on raspberry pi just make sure you have all the libraries installed (as I don't know if we can download packages on raspberry the same way we do one our PCs). Second thing this program takes input from microphone so you have to configure a mic on raspberry which you can follow up from any UA-cam video. All the best!
This proposed method does not work in general - it was too good to be true. It works only for the recorded segment and does not work in general - I did think it was easy to build a wake word detection system. I wasted time trying it out
Hello, many people are facing similar issues. In my case it was working so i thought to share my work with everyone. I am thinking to re-create this project but this time to make it much more robust that will work for everyone. I am disappointed myself that it didn't work out for everyone.
@@PritishMishra Thanks for the reply. I hope you could update the video or make it clear, I could have saved 12 hours of time watching and working and realizing it - it could have save me whole days of frustration. Unfortunately Google is so bad, it shows your video as first recommended video for wake word. You have taken good effort and I am sure your intention is to help and save time for others. I wish Google could show rating and ranking on videos so I could have saved over one day of working through that got me no where. Thanks for effort to share
Heyy, have problem with the training.py! I try to install pcm.pcm module and with pip everything works, but when I start the programm I get error (ModuleNotFoundError: No module named 'pcm.pcm') Pleas help me ;(
I used 100 samples of me saying the wake word, then I used 100steps of background sounds, and the result is that I get continuously positive matches, even if i say nothing.
Part 2: ua-cam.com/video/gl1lhplZmaI/v-deo.html
If the negative is just background sound, how do you know the model just didn't become good at detecting when someone says any word?
Waiting for second part🤩🤩
Nice Video , waiting for the next part .
Model Classification Report:
precision recall f1-score support
0 0.92 1.00 0.96 33
1 1.00 0.90 0.95 31
accuracy 0.95 64
macro avg 0.96 0.95 0.95 64
weighted avg 0.96 0.95 0.95 64
This is your model when training your dataset. Mine is not like that or is it because I got the wrong data. I follow your step 1 instructions when getting data
Wating for next video
i like your channel, nice intro ! 👍
What’s the point of setting 2 channels while recording? Are you using stereo microphone?
Hello ! Great tutorial video ! How could I use this in order for it to work properly using 2 or more different wakewords? What should I change?
Принцип dry типа не проходили или это дурное влияние Индии?
hey bro literally greattt !!!
Hello I'm trying your all step but the tensor flow is not working I don't know how I've stopped training what can I do
I am using intel mini pc ubuntu os
please create a new video on this topic using cnn as this model is not working accurately most of the time..
Sure. I am thinking the same.
How can I avoid false detection? The program is triggered when there is no wake word as well. Please help me soon. Thanks.
There was similar issues with me. All i did is train it on more voice data and add more background audios with variations. And it improved with each trial.
@@PritishMishra Okay, thanks.
Waiting for the next part....
I have been in a hurry to try this so I had launched training.py but it says missing module PCM but I have installed it using pip install pcm but again this message is coming.Any advice
@@SaloSojaEdwin github.com/pritishmishra703/cmtoolkit.git and see video: ua-cam.com/video/oZ6tFvWgfYE/v-deo.html
@@SaloSojaEdwin You can Directly install this package by running this command:-
pip install git+github.com/pritishmishra703/cmtoolkit.git
@@PritishMishra Thanks for the reply, I have fixed that. And trained, But when I execute prediction.py I am getting many false positive
@@SaloSojaEdwin Yeah in the starting I was getting many false positives too... In the part 2 I will tell how to avoid false positives almost completely! Will be uploading tomorrow or day after tomorrow...
how can i use this and make a personal assistant using wake word detection
I have another video coming up soon, and it's on a similar topic, please stay tuned :-)
please help me i want to add second hortword detection ,which should i edit in ur file pls help me faster
Just record wake word with multiple words. For example, record 100 audio with first wake word and then 100 audio with the second wake word.
I don't know how well this will work as I have not tried this myself.
Nice one🤩🤩🤩🤩
Hey man, wanted to ask if this would work on a raspbery pi?
Yeah definitely, it is a python script you can run the same on raspberry pi just make sure you have all the libraries installed (as I don't know if we can download packages on raspberry the same way we do one our PCs). Second thing this program takes input from microphone so you have to configure a mic on raspberry which you can follow up from any UA-cam video. All the best!
This proposed method does not work in general - it was too good to be true. It works only for the recorded segment and does not work in general - I did think it was easy to build a wake word detection system. I wasted time trying it out
Hello, many people are facing similar issues. In my case it was working so i thought to share my work with everyone.
I am thinking to re-create this project but this time to make it much more robust that will work for everyone.
I am disappointed myself that it didn't work out for everyone.
@@PritishMishra Thanks for the reply. I hope you could update the video or make it clear, I could have saved 12 hours of time watching and working and realizing it - it could have save me whole days of frustration. Unfortunately Google is so bad, it shows your video as first recommended video for wake word.
You have taken good effort and I am sure your intention is to help and save time for others.
I wish Google could show rating and ranking on videos so I could have saved over one day of working through that got me no where. Thanks for effort to share
Heyy, have problem with the training.py! I try to install pcm.pcm module and with pip everything works, but when I start the programm I get error (ModuleNotFoundError: No module named 'pcm.pcm') Pleas help me ;(
Run this in terminal:
pip install git+github.com/pritishmishra703/cmtoolkit.git
And run the training.py again it should work now!
@@PritishMishra It doesn't, I keep getting the Module not found for the PCM.
Nevermind, I just copied the file.^^
I used 100 samples of me saying the wake word, then I used 100steps of background sounds, and the result is that I get continuously positive matches, even if i say nothing.
After PreprocessingData. My data show:
1.0
Model Classification Report:
precision recall f1-score support
0 1.00 1.00 1.00 33
1 1.00 1.00 1.00 31
accuracy 1.00 64
macro avg 1.00 1.00 1.00 64
weighted avg 1.00 1.00 1.00 64.
What is reason?
I think you have ran training.py
@@PritishMishra I want to ask, the more wakework files, the better the model, right?
@@HangNguyen-kx4iq Yes! More the data more better will be your Model!
How can I enable it to wake using anyones voice
You will have to train the model on very different voices then.
Great